FREESIDE
freeside.io
Report CurrentsMarketsCryptoTech & AIPoliticsGeopoliticsHealth
Daily ecosystem report

The market as a system of participants, not a price.

Every day we record who was forced out of their position, where the visible crowd turns out to be a single algorithm, how far the advertised book diverges from what actually traded, and what it costs an ordinary participant to take part at all.

Window: 1 Sep 2026, 22:17 UTC → 2 Sep 2026, 22:17 UTC
01 — Forced flow

Who was forced

Every position below was closed by the exchange, not by the person who opened it. Forced exits are the one moment a participant has no choice and no discretion — which makes them the cleanest record the market produces of who was actually offside.

$25.42M
Forcibly closed
8,854
Separate liquidations
270
Instruments touched

Longs accounted for $16.17M across 5,289 events on 270 instruments; shorts $9.25M across 3,565 events on 185 instruments. The single largest forced exit was $495,658 in SNDK-USDT (a long).

543 of these events cannot be valued in dollars — they are delisted instruments with no contract multiplier on record. They are counted in the event totals above and excluded from every dollar figure. We would rather show you the hole than quietly close it.

InstrumentEventsForced value
ETH-USDT 705 $4.06M
BTC-USDT 459 $2.89M
SOL-USDT 330 $2.19M
XAU-USDT 235 $1.90M
SNDK-USDT 414 $1.88M
XRP-USDT 182 $1.40M
UNI-USDT 483 $1.09M
ARB-USDT 627 $829K
8,854 forced exits recorded from OKX across 270 instruments. 543 could not be valued in USD (delisted instruments, no contract multiplier on record) and are counted but excluded from every dollar total below.
02 — Book structure

Where there was no crowd

An order book looks like a crowd: many levels, many sizes, apparent depth. Sometimes it is not a crowd at all. When most levels on a side carry the same clip size, you are not looking at many participants — you are looking at one market maker's ladder wearing the costume of a market.

2 / 60
Books that were one algorithm
3%
Of the books we sampled

We apply a deliberately conservative rule: a side counts as a quoting grid only when the single most common clip size accounts for 60% or more of its levels, across at least six priced levels. Genuine multi-participant books sit far below that line — the largest instruments in this sample typically show around 10%.

Worked example · PIEVERSE-USDT

70% of the top 10 bid levels carried the identical clip size of 77.62 units. Real crowds do not agree on a size to four significant figures six times in a row — that is one participant's quoting ladder, refreshed by algorithm.

$4,813Total bid depth, top 10 levels
5.4 bpSpread to cross

The sample is stratified, drawn evenly across the activity range of the 232 spot instruments we track, rather than taken from the busiest names only. That matters: the most heavily traded books are the ones most likely to contain a genuine crowd, so sampling those alone would answer an easier question than the one asked.

This is a point-in-time reading, taken once when this report was built — not a full-day statistic. The detector runs against the live ladder and its output is not retained, so we can honestly report what the books looked like at that moment and nothing more.

Stratified sample of 60 live KuCoin spot books, drawn evenly across the activity range of the 232 spot instruments we track.
03 — Advertised vs delivered

Quotes versus fills

Quotes are an advertisement. Fills are a receipt. The gap between them is the most under-reported number in this market.

$17K
Median depth advertised within 1% of mid
$51
Median value actually executed per minute
194:1
Advertised versus delivered

For the median instrument, the book displays about $17K of resting size within one percent of the mid price, while roughly $51 changes hands in a typical minute. These measure different things on purpose — depth is a stock standing at an instant, volume is a flow over a minute — and we are not claiming they should match. The point is the order of magnitude between them. Displayed liquidity is overwhelmingly liquidity that never trades.

InstrumentAdvertisedExecuted / minRatio
2Z-USDT $28K $1 25,497:1
XPL-USDT $125K $9 14,356:1
GRASS-USDT $146K $12 12,051:1
XDC-USDT $207K $28 7,506:1
XEC-USDT $23K $4 5,910:1
Widest gaps among instruments advertising at least $20K of depth.
224 KuCoin spot instruments had both a recorded order book and a recorded tape for at least 60 minutes of the window. Futures contracts are excluded: their book sizes are denominated in contracts rather than base units, so a like-for-like dollar comparison is not available for them.
04 — Economics

The cost to participate

Participation is not free, and the fee is not the interesting part — the interesting part is how the fee compares to the movement available to pay it.

0.064%
Median instrument's entire range, per minute
0.12%
Cost of one futures round trip (taker)
69%
Of traded minutes moved less than that

In 68.9% of all traded minutes, the distance from the high to the low — the entire range, perfectly timed, bought at the bottom and sold at the top — was smaller than the 0.12% it costs to go in and out once on futures. On spot, at 0.20% the round trip, that rises to 82.7%.

This is not because the market is quiet. Over the same window the median instrument travelled 5.6% from its high to its low across the day, and not one of the 406 instruments we measured had a daily range smaller than the round trip. The movement is there. It is simply not there at the timescale most participants trade. Shorten the holding period far enough and the rake stops being a cost of doing business and becomes the business.

Fee tiers are the ones we are actually charged on the venues we record: 6 bp per side taker on futures (2 bp maker), 10 bp per side on spot. Fees vary by venue and volume tier; yours may differ.

361,837 instrument-minutes with at least one executed trade, across KuCoin spot and futures. Unlike the dollar figures above, a high-to-low range is a percentage and so is unaffected by the contract-vs-base unit difference between the two; measured separately the two agree (spot 70.7%, futures 67.2% of minutes below the futures round trip).
05 — Reported observations

What was worth saying

Unusual events picked out of the streams we record, scored against their own history. Each one was reviewed before publication. Nothing here concerns our own trading.

Not trading advice. This is an automated observation of public market data — what happened, not what you should do about it.

Forced liquidations, last 24h: $25.82M across 299 symbols

8,643 forced liquidations were recorded in the window ending 2026-09-02 20:00 UTC.

Total liquidated notional: $25.82M, computed from the 95% of liquidations with a derivable USD value.

Of that, $16.68M was liquidated longs (forced selling) and $9.14M liquidated shorts (forced buying).

Largest single liquidation: $999.82 in AAVE-USDT.

Most liquidated notional: ETH-USDT, BTC-USDT, SOL-USDT, XAU-USDT, SNDK-USDT.

We record OKX and KuCoin. This is not the whole market. Coverage for the weakest figure here: 95% of liquidations in the window with a computable USD notional. Insufficient coverage: 'total_notional_usd' observed 95% of liquidations in the window with a computable USD notional (floor 95%).

How this was measured

Source: liquidation_event — OKX public liquidation feed, collected by Tape.

Significance: not scored.

  • OKX perpetual swaps only — we do not record liquidations from Binance, Bybit, or any other venue, so market-wide totals are larger than anything stated here
  • forced liquidations only — voluntary closes, and positions reduced before liquidation, are invisible to this feed
  • a scheduled daily ledger: it states what we recorded and does not claim the day was unusual in either direction

https://freeside.io/ecosystem/2026-09-02#obs-34877

Provenance

Report built
2 Sep 2026, 22:17 UTC
Window
24 hours to 2 Sep 2026, 22:17 UTC
Forced-flow feed
OKX perpetual swaps — 301 instruments with events
Book & tape feed
KuCoin spot — 330 instruments with recorded book, 414 with recorded tape
Method
Aggregated directly from our own recorded data. No third-party analytics.

What we do not see

  • Two venues only. Forced flow is OKX perpetual swaps; book and tape are KuCoin spot. Binance, Bybit and the rest are outside our recording, so no figure here is a whole-market total.
  • No over-the-counter or internalised flow. Large trades negotiated off-book never reach a public tape, and they are precisely the trades that move things.
  • Order books are sampled, not streamed to disk. We keep per-minute depth metrics, not every ladder update, so sub-minute book behaviour is not reconstructable after the fact.
  • Cross-venue identity is unresolved. The same instrument on two exchanges is two records to us; we do not net them into one participant view.
  • Section 02 is a point-in-time sample of the most-active instruments, not a census of every book across the day.